24.  Ledgers are lists of records where transactions are recorded once and cannot be subsequently updated. This means that any changes must be recorded as new transactions (book-keeping entries). Digital ledgers may be stored as a database, also known as a journal database. Each record can be read many times but written only once. The term ledger comes from accounting where entries, once written into a ledger (accounting journal), cannot be changed. A blockchain database is a ledger because it uses hashes to ensure that none of the data it contains has ever been changed.

25.  A blockchain ledger database is described as being distributed because there are multiple copies kept on different nodes. The multiple copies are updated with new data blocks in a coordinated way that ensures they remain consistent, using a consensus algorithm of which there are different types.

26.  In summary, the content and sequence of the data blocks in a blockchain are determined by a consensus of the participating nodes and each block contains a fingerprint (hash) that can be used to recursively verify the content of all previous blocks.

27.  Each block of data written to a blockchain ledger contains at least one record of a transaction, although most blocks contain many records of transactions. A simple example of a transaction would be “debit one coin from account A, and credit one coin to account B”, although many other kinds of transactions are possible. Some blockchains support a limited sub-set of transactions (operations or algorithms) such as this simple double-entry bookkeeping operation. Some blockchains support a much wider set of transactions covering any solvable algorithm (i.e. a Turing-complete computer programming language⁴). These types of transactions are variously called smart contracts, chaincode, transaction families, or other equivalent terms. In summary, all blockchains support a variety of data operations on their chains, but not all blockchains support Turing-complete transaction languages.

28.  Blockchains implement two kinds of cryptographic technology: hash functions and public/private key cryptography. Hash functions are used to construct the fundamental proof that links each block to the rest of the chain before it. Hashes, in a different context, can also be used to provide proof of validity for data that is referenced by blocks and they are used in Proof-of-Work consensus algorithms where a hash with a specified number of leading zeros serves as the “difficult problem” that nodes must solve in order to reach consensus.

29.  Public/private key cryptography is used for identifying parties to a transaction and controlling access to data. An analogy is email, where the public key is your email address which others can use to send messages to you, and the private key is your password which gives access to the private material, which is your messages. So, on a blockchain, a public key can be used, for example, to implement a transaction that sends a document or a payment to a party, but only the party with the private key can access those documents or payments after they are sent.

30.  There are various consensus algorithms used by different blockchain systems. For example, Bitcoin, a public blockchain, uses Proof of Work algorithms which allow miners to recover the cost of computationally expensive work in exchange for transaction fees and these fees also provide a way to initially put coins into circulation. Permissioned ledgers use a consortium of collectively trusted, but not necessarily individually trusted, nodes to agree on the output of a consensus process—which is generally cheaper and faster than Bitcoin’s Proof of Work. All consensus processes require a mechanism to settle disputes, or uncertainty, about which block should be written next. Most of these mechanisms are based upon using the block, which is agreed upon by more than 50% of the nodes. A more detailed description of public and permissioned blockchains can be found below.

31.  The nature of the consensus mechanism determines some key characteristics of a blockchain system. For example, mining the creation of blocks has deliberately been made expensive. This protects the blockchain by making the cost of capturing more than 50% of the nodes—the number needed to approve a block, and thus to manipulate the blockchain—prohibitively expensive. To compensate for this cost, miners are rewarded both an amount of Bitcoin for each block they create and fees for each transaction written to the blockchain⁵. Each block has a size limit and transaction costs are determined on a free-market basis, so the more transactions are requested, the more the price increases for each transaction. This is necessary for the Bitcoin economic operating model, which seeks to obtain an honest consensus in an unregulated market of potentially anonymous and economically rational operators (i.e. operators who might, being anonymous, and having no costs for doing so, steal assets). As an additional incentive, if a node/miner does not accept the block voted on by over 50% of the other nodes, it is effectively kicked off the blockchain, thus losing the possibility of earning future Bitcoins and transaction fees. Consequently, Bitcoin has extremely low bandwidth due to the cost of generating blocks with transactions taking on average 10 minutes to be confirmed. In addition, its very large number of nodes and users, generating large amounts of data, together with its block-size limits, makes storing data on the Bitcoin blockchain expensive as well as being inefficient.

32.  Given the duplication of information across all nodes on a blockchain, it is generally inefficient to store significant amounts of data on blockchains. Bitcoin still supports many billions of US dollars’ worth of Bitcoin and other high-value transactions, but its speed and volume limitations make it unsuitable for many enterprise applications and the direct implementation of small-value transactions.

33.  Permissioned ledgers strike a different balance between bandwidth, capacity and trustworthiness. For example, because they have more control over who participates, permissioned ledgers can use other consensus mechanisms—even if some of them are somewhat less robust than the Proof of Work used by Bitcoin. For example, there are consensus mechanisms based on the amount a node has invested in a network (called Proof of Stake), or where a consensus by a subset of nodes is verified by a larger group.

34.  In addition, there is a great deal of research by foundations, universities and companies looking to identify and test other consensus mechanisms. Some of these alternative consensus mechanisms will allow ledgers to support hundreds or even thousands of transactions per second, rather than an average of one new block per 10 minutes, as with Bitcoin. There is also research going into the maintenance and accessing of data on petabyte-scale (i.e. truly gigantic) databases.

35.  When consensus is reached, which includes agreeing that a block contains legitimate data, and that it is the block that should be written next, each node adds the agreed block to their local copy of the ledger. In this way, all nodes maintain an identical copy of the ledger each time a block is written. This is proven by the next block to be written, because it will contain a hash of the block before it.

36.  Recall that a hash is a one-way function that produces a unique fingerprint of selected data. Also note that a hash function produces a fixed-size fingerprint regardless of the amount of data being hashed. As a result, there is no way to know from looking at the hash if the data was a single, small document or a database holding many billions of records.

37.  Each block in a blockchain contains some transaction data plus the hash of the previous block, which is always the same size no matter how much data it represents. Given a consensus that this new block forms part of the chain, it is possible to verify the previous block from its hash—and from the previous block, the block before it, and so on all the way to the first or genesis block in the chain. The hash of the previous block is said to be anchored in the subsequent block.

38.  Tampering with the contents of any block in the chain will change the hash of that block, which will change the hash of the block after it, and so on for every subsequent block in the chain. If this occurs then the tampering is easily detectable by any node, and the consensus algorithms will prevent new blocks from being written to the chain because the hashes don’t match.

39.  This characteristic is the origin of the word “chain” in “blockchain” because each block is anchored to the previous block and proves the existence of all the data it references going back to the first “block” of data in the “chain”.

40.  Public ledgers can be read by anyone. They are also permissionless because anyone can participate and utilize the consensus mechanisms without needing permission to do so and without depending on a regulator to enforce acceptable behaviour. Bitcoin, Ether and a range of other cryptocurrencies with market capitalizations going up to 59 billion USD⁶ operate this way, allowing any transaction that is logically valid between any parties on the network, including anonymous and pseudonymous parties.

41.  One of the fears about blockchain technology is that, if a malevolent actor were to control a majority of the nodes, then they could decide to reach a consensus in contradiction of the interests of other stakeholders. This threat is called a Sybil attack in cryptographic literature. A successful Sybil attack on a public blockchain cryptocurrency could result in a catastrophic redistribution of assets and/or double spending. Public blockchain ledgers are designed to operate according to rules that do not require governance or regulatory mechanisms to intervene in order to prevent antisocial transactions, because those mechanisms might themselves be exploited for antisocial outcomes—for example, if a governance mechanism were to be hacked by a third party or abused by a trusted regulator. Public blockchains operate with absolute trust in their algorithms and are designed to avoid any need to trust any counterparties. This is why public blockchains are sometimes referred to as being trustless.

42.  Public ledgers typically compromise other aspects of performance in order to achieve a strong resistance to Sybil attacks. They also rely on the transparency of the public ledger, and on the transparency of the open-source software involved.

43.  Like conventional databases, the contents of a private blockchain ledger may be a guarded secret that is only available to selected users, and node operators, through a role-based access control mechanism. Likewise, a private blockchain can be set up so that everyone can read the data, but only designated nodes can add new data. This can also be done on a public database using smart contracts, however, authorities might be concerned that there is a greater security risk since anyone who wants to could see (and try to hack) the smart contracts in question. Such a database might be desirable for official records such as land deeds, licences, certificates, etc. Unlike a traditional database, a private blockchain ledger is immutable (i.e. cannot be updated) and transactions are verified by a consensus mechanism that is established by the network operators.

44.  Private ledger technology is typically applied in enterprise use cases where immutable transactions are required that can be verified by a closed community of nodes. These nodes may be independent of parties to the transactions on the blockchain and may be subject to oversight and governance that is not possible, or considered desirable, in a permissionless, public blockchain system.

45.  Permissioned ledgers operate with a different threat model to the public ledgers. The operators of permissioned ledger nodes are not anonymous; they are subject to some kind of governance controls and are collectively trusted by the users. Antisocial behaviour by a node or participant could result in that party being evicted from the network and their transactions blocked. The expectation of users of a permissioned ledger is that the operators will intervene in antisocial behaviour but not commit antisocial behaviour themselves.

46.  On permissioned ledgers, the level of security, and so the confidence users can have in the immutability of the data, varies depending upon the rules established for that permissioned ledger, including its consensus mechanism. Permissioned ledgers can also create a false sense of security because only trusted participants are allowed to maintain nodes and participate in verification. At the same time, even trusted participants can become untrustworthy upon being hacked; permissioned ledgers with single points of failure are also vulnerable should anything happen to that single point, and poorly tested smart contracts can create bad consequences for participants—even if no harm was originally intended—especially if the blockchain network does not have adequate controls in place.

47.  Today, many different blockchains exist and in the future, there will be even more. Already, a supply chain transaction, from beginning to end, could involve writing or reading data from multiple blockchains. For example, an exporter might need to use a bank blockchain, one blockchain per transportation mode, a blockchain used for traceability by the importer and one or more used by regulatory authorities. In addition, it is easy to foresee an increasing need for the exchange of information and the implementation of transactions across blockchains (i.e. interledger).

48.  Blockchains have the possibility to reference data outside of that blockchain. This includes data in other blockchains as well as from non-blockchain systems. There are two broad categories of external data references that can occur in a blockchain system: linked data and blockchain-spanning transactions.

49.  Linked data uses hashes and may also use digital identifiers and public key cryptography. This will work as long as the rules are used consistently across the blockchain and the system(s) the linked data is stored on. This implies that the more standardized the use of public-key cryptography, the easier and less expensive it will be to link data—and the same can be said for the semantics defining the data. The use of common semantics (i.e. data definitions) greatly simplifies the job of interpreting data from different sources and the UN/CEFACT Core Components Library is a very complete library of trade-related semantics which can be used in this context.

50.  Blockchain references which point to external data (also known as anchors) can also contain information, such as hashes, to be used to prove the existence or unchanged nature of the data referenced. This is different from a hyperlink or Uniform Resource Locator (URL) on the Internet where the information at an address may change depending on the time it is accessed. For example, if you click on a link on a television news website, which changes on a regular basis as it is updated, what you find tomorrow may be different from what you find today. With a blockchain anchor data link, the information in the blockchain is a guarantee (proof of existence) that the data being pointed to has not been changed.

51.  In addition to linking data between two blockchain systems (cross-chain references) and pointing to data that may be used by a smart contract (for example a test certificate) in a more standard database, linked data can also be used to incorporate off-chain big data into a space-constrained blockchain system. Supplementary data can either be in public/open distributed data systems such as the InterPlanetary File System (IPFS)—an open, content-addressable memory that uses standard internet protocols—or it may reference data in private databases that are selectively available to permissioned ledger users. With private off-chain or cross-chain references, it is possible for network operators to know that some data exists, but to have their access limited by additional controls. This can be very interesting from a privacy standpoint as it is possible to access data in order to know that, for example, someone is over 21 without giving their age, or that they live in London, without giving their address.

52.  These sources of external data are sometimes called oracles which are described in more detail below.

53.  Interledger (blockchain-spanning) transactions use cross-chain references and smart contracts (see description below) on both blockchains that interact in a coordinated way. This is an emerging field, however there are mechanisms that already exist and are in use. These are primarily focused on exchanging value (i.e. digital assets) between ledgers, for example Ripple Interledger and the Lightning Network.

54.  Smart contracts are self-executing computer programs that encode business logic. They execute when pre-defined conditions are met. In other words, their execution is not launched, or at least not directly, by human intervention. These can be as simple as “transfer specific amount of asset from account X to account Y.” Smart contracts are based on the conditional If-This-Then-That (IFTTT) model where some activity is automatically executed when certain conditions are met. These conditions can be a certain period of time, a specific value (for example the price of some asset, such as stock) or a specific event such as the delivery of ordered goods to a customer.

55.  Smart contracts offer several benefits:

• Improved security and predictability because they eliminate the human element and potential contract breaches intentionally or unintentionally caused by human action;

• Transparency because the code of a smart contract can be public and visible, anyone can review it and predict how transactions under a given contract will behave; and

• Simplified programming for systems that need to accept, match and then act upon data from a wide variety of parties, many of whom may be unknown.

56.  One example of a smart contract explained in everyday language could be:

• Precondition: when I deposit a certain amount of cryptocurrency and the other party deposits a certain amount of FIAT currency;

• Condition: if the amounts are equal according to the current exchange ratio; or

• Action: then currencies are exchanged between involved parties’ accounts.

57.  Another example could be when renting a car; the rental agency could require that an advance currency deposit be made on a blockchain. The amount would then only be released to the rental agency after the renter confirms that he/she received the car’s keys. This way smart contracts can prevent scams based on advance payments and create an additional layer of insurance.

58.  Because smart contracts are basically small programs, they can be developed and customized for many situations, making them potentially powerful tools for business.

59.  The primary function of oracles is to provide secure and trustworthy data to a blockchain smart contract. Smart contracts then look at this data to see if it meets the conditions defined in the smart contract’s code and, if this is the case, the contract automatically executes.

60.  The key words here are “secure and trustworthy data”. Blockchains cannot, and should not, store large amounts of data, so information needs to be submitted to the blockchain via an oracle. This makes the oracle (just like user interfaces) a weak point in the security and integrity of a blockchain. It is also where the old adage of “garbage in—garbage out” come into play (although in the case of blockchains it may be garbage in—garbage forever). Therefore, it is very important in blockchain-based applications to carefully design the process for obtaining the data used by oracles as well as their interfaces with blockchains to ensure the quality and integrity of the data and related processes.

61.  The Internet of Things (IoT) refers to sensors and small computing devices or chips embedded in physical objects which communicate via the Internet. These communications can be with one another, with larger computers and computing systems and even with humans—for example modern security systems that notify a homeowner if they detect motion in the owner’s home and connect the owner with the video camera in his or her living room.

62.  IoT devices can collect a wide variety of data. Examples of information related to trade and transport communicated by IoT devices include truck or container location and movements via GPS coordinates; the opening and closing of container doors; container temperatures; external shocks to containers/pallets/products; and, for very expensive items such as some pharmaceuticals or luxury goods, the tracking or identification of individual packages or products.

63.  IoT devices can be a useful way to capture data that is analysed by other systems that then supply the analyses’ results to a blockchain (i.e. systems that are blockchain oracles), or they can be oracles themselves by providing data directly to a blockchain. Nonetheless, IoT devices tend not to be used directly as oracles because of security concerns, and because systems that are connected to tens of thousands of IoT devices might be overwhelmed by data volumes. Also, writing constant data readings to a blockchain could be expensive for those networks where every time you write data you have to pay a small amount. As a result, data from the IoT is often filtered so that only data which goes outside of defined ranges is communicated, or the data is communicated as a total set of readings at the end of a process.

64.  A classic example of the use of IoT data by a blockchain is insurance for temperature-sensitive goods (i.e. fruit that is supposed to be kept at between 4 and 15 degrees Celsius during shipment). During shipment an IoT device in a container records that the fruit was kept at 0 degrees Celsius for 2 entire days. This information is given to the smart contract which notifies the insurance company that a payment should be made to the exporter to compensate for the goods destroyed by the excessively low temperature and that payment is automatically made by the smart contract without any further intervention by either the importer, the exporter or the transport company. This significantly decreases the cost for insurance companies of processing claims because they do not have to reconcile information submitted by the shipper/exporter with the insurance policy, evaluate the truth of the insurance claim (the IoT data provided the proof) and then request payment. In addition, it reduces the costs for the shipper/exporter as they do not have to undertake any further documentation of the problem which occurred, and they receive their insurance payment more quickly.